Navegacion en Tortugas

7/29/2019 Navegacion en Tortugas

1/9

Anim Cogn (2009) 12:779787

DOI 10.1007/s10071-009-0237-9

123

ORIGINAL PAPER

Visual and response-based navigation

in the tortoise (Geochelone carbonaria)

Anna Wilkinson Sacha Coward GeoVrey Hall

Received: 10 November 2008 / Revised: 4 May 2009 / Accepted: 6 May 2009 / Published online: 23 May 2009

Springer-Verlag 2009

Abstract Much research has investigated spatial cogni-

tion in mammals and birds. Evidence suggests that the hip-pocampus plays a critical role in this; however, reptiles do

not possess a hippocampus. It has been proposed that the

reptilian medial cortex plays a similar role, yet little behav-

ioral research has directly investigated this. Consequently,

this study examined the role of extramaze cues in spatial

navigation by the red-footed tortoise (Geochelone carbona-

ria) using an eight-arm radial maze. In Experiment 1 the

maze was surrounded by a black curtain on which geomet-

rical shapes were attached. After the tortoise reached

above-chance performance we introduced test sessions in

which the cues were removed. Performance was unaVected

by cue removal. The tortoise appeared to have developed a

turn-by-one-arm strategy. In a second experiment the

curtain was removed and the tortoise was allowed access to

a rich-cue environment. The use of the turn-by-one-arm

strategy was signiWcantly reduced and the tortoise appeared

to be using the extramaze cues to navigate around the appa-

ratus. This type of response-based strategy, and the speciWc

contexts in which it was used, has not been observed in

mammals and birds, suggesting that the mechanisms served

by the reptilian medial cortex do not parallel exactly those

of the hippocampus.

Keywords Reptile Spatial learning Radial maze

Extramaze cues Response strategy

Introduction

Studies of navigation in mammals and birds have shown

that they are able to use a range of strategies to navigate to a

speciWc goal. These include the use of a single salient cue

(a beacon), path integration, learning a sequence of

responses, and creating a spatial representation of environ-

mental cues (a cognitive map). In the latter case the goal is

deWned by its spatial relation to a number of diVerent land-

marks; this may be seen as highly adaptive as the removal

of any single landmark does not necessarily disrupt naviga-

tion. This behavior is thought to be dependent on the hippo-

campus (OKeefe and Nadel 1978). Evidence that

demonstrates the use of this navigational system has been

found in a variety of mammals and birds (see OKeefe and

Nadel 1978 for a summary), and also, more recently, in Wsh

(for a review see Broglio et al. 2005). As such it is likely

that reptiles also possess a similar system, there is some

evidence to support this (Lpez et al. 2000, 2001, 2003);

however, its existence in this group is much less investi-

gated and the Wndings are not clear cut (Day et al. 1999,

2001).

The study of spatial behavior in reptiles has a long history

(e.g., Tinklepaugh 1932; for a review see Burghardt 1977),

but little is known about the mechanisms underlying their

navigational ability. It is still not clear whether reptiles are

capable of the forms of spatial learning that are seen in

mammals and birds. For example, Holtzman et al. (1999)

found that corn snakes (Elaphe guttata guttata) could rap-

idly learn the position of a hidden goal in an open Weld task.

Their results (though not designed to test the underlying

A. Wilkinson S. Coward G. Hall

Department of Psychology, University of York, York

YO10 5DD, UK

A. Wilkinson (&)

Department of Neurobiology and Cognition, University of Vienna,

Althanstr. 14, 1090 Vienna, Austria

e-mail: [email protected]


2/9

780 Anim Cogn (2009) 12:779787

123

mechanisms) suggested that the snakes used a beacon to

navigate to the goal. However, very diVerent results have

been found in lizards. Day et al. (1999) examined the diVer-

ent spatial abilities of two closely related lizard species

whose foraging strategies diVered. They found that neither

species preferentially attended to either distal or local cues

to navigate. The lizards (of both species) rarely approached

the goal directly, suggesting that they found it through trialand error searching. This led the authors to conclude that, if

spatial memory exists in reptiles it is fundamentally diVerent

from that observed in mammals and birds (Day et al. 1999).

Quite diVerent results have been found in the study of

spatial learning in chelonian (turtles, terrapins, and tor-

toises). Lpez et al. (2000) showed that the terrapin (Pseu-

demys scripta) was capable of learning to approach a given

location in a T maze, regardless of the starting position. The

terrapins behavior appeared to be based on extramaze cues

and removal of these (using a curtain to block oV parts of

the experimental room) disrupted performance. The authors

suggested that the terrapins were navigating using a cogni-

tive map of the sort that is postulated in mammals and

birds. Later Wndings revealed a Xexibility in the use of this

strategy. Lpez et al. (2001) required one group of terrapins

to navigate to a speciWc goal solely on the basis of distal

cues; a second group had the same distal cues but also had a

beacon (a large salient cue) that was located close to the

goal. Both groups learned to successfully navigate to the

goal, and probe trials revealed that only the group with just

the distal cues appeared to make use of a cognitive map-

like representation to do so. The beacon group did not use

the distal cues and, instead, used the single beacon to locate

the goal. This pattern of results suggests that reptilian learn-

ing and memory capabilities (in terrapins at least) may

closely parallel those observed in mammals and birds.

Further evidence to support this conclusion comes from

studies of the cerebral basis of spatial navigation in terra-

pins. It has been suggested that the reptilian medial cortex

serves a parallel function to the mammalian hippocampus

(e.g., MacPhail 1982; Lpez et al. 2003). Experiments by

Lpez et al. (2003) revealed that lesions to the medial cor-

tex of terrapins caused a change in performance on an open

Weld task when the animals had to use distal landmarks to

navigate to a single goal. After a number of post-operative

training sessions the lesioned animals were able to learn to

navigate to the correct location, with performance equaling

their pre-operative level (and the level shown by sham-

operated animals). Probe trials revealed that the lesioned

terrapins used a modiWed beacon strategy to locate the goal,

whereas the sham-operated animals used a map-like strat-

egy. These results closely parallel those found in mammals

and birds and further suggest that the medial cortex per-

forms a similar function to the mammalian/avian hippo-

campus.

This set of results suggests that reptiles (or at least terra-

pins) are capable of at least some basic spatial tasks, and

that the medial cortex plays an important role in this. How-

ever, little is known about the limits of their spatial ability.

Previous studies have tested the ability to navigate to one

speciWc area in which the animals were rewarded; however,

a more natural task would examine their ability to remem-

ber numerous places that they have previously visited toensure that they do not return to an area in which a food

source has already been depleted. The classic task for this

test is the radial arm maze.

In a recent investigation, Wilkinson et al. (2007) exam-

ined the ability of a single red-footed tortoise (Geochelone

carbonaria) to navigate in an eight-arm radial maze. The

tortoise showed reliable, above-chance performance, pref-

erentially choosing baited arms rather than returning to

arms previously visited within a trial. Tests ruled out the

use of olfactory cues from either the bait (strawberries), or

from the avoidance of odor trails. This suggests that, in

spite of diVerences in brain structure, the tortoise showed

spatial learning abilities comparable to those observed in

mammals.

The cues that controlled this behavior remain to be deter-

mined. As suggested by Lpez et al. (2000, 2001) for terra-

pins, it is possible that the tortoise is able to identify spatial

locations by the conWguration of visible extramaze cues that

deWne them. But diVerences in the behavioral ecology of ter-

rapins and tortoises suggest that it may be unwise to assume

that the same mechanisms are used by both. Research with

corvids has revealed the importance of behavioral ecology

in spatial navigation (for a review see Balda and Kamil

2006). Very diVerent spatial abilities have been observed in

closely related species, and species that are phylogenetically

disparate but inhabit similar ecological niches have devel-

oped similar (highly specialized) spatial memory systems

(Kamil et al. 1994). The behavioral ecology of the red-

footed tortoise is quite diVerent from that of Lpezs terra-

pins; the former actively forage for fallen fruit and Xowers

(Strong 2005; Strong and Fragoso 2006), whereas the terra-

pin is largely carnivorous whilst young, but as an adult eats

largely plant matter (Hart 1983). Moreover, the habitat

occupied by the two study species is very diVerent. The red-

footed tortoise is terrestrial and lives on the margins of trop-

ical forests (Strong 2005), whereas the terrapin of Lpezs

experiments inhabits areas surrounding freshwater lakes,

rivers, and ponds. These diVerences, in conjunction with the

diVerences found within the reptile group itself makes it

possible that quite diVerent spatial abilities are present in

these closely related species and that the performance of the

tortoise may more closely approximate that of a mammal

inhabiting a similar ecological niche.

In order to resolve this issue, it is necessary to conduct a

study in which the role of extramaze cues in controlling


3/9

Anim Cogn (2009) 12:779787 781

123

navigation in the tortoise can be investigated directly. In

order to do this we returned to the basic procedure used by

Wilkinson et al. (2007) in which the tortoise was trained in

an eight-arm radial maze, but changed the nature of the

cues available. In the previous study the maze was open to

the room. In the present study (Experiment 1) we sur-

rounded the maze with a curtain, but Wxed to this a set of

cues (simple geometrical shapes) that could be manipulatedin a post-acquisition test, to determine the role of extramaze

cues in controlling the animals behavior.

Experiment 1

In this task the animal was initially trained in an eight-arm

radial maze. All arms were baited with food and we

expected, on the basis of previous Wndings (Wilkinson et al.

2007) that the tortoise would learn, in the course of a trial,

to visit each arm just once. In our previous experiment the

maze was open to the room. In this experiment it was sur-

rounded by a curtain to which visual cues were attached.

These cues were removed for the post-acquisition test

phase.

Method

Subject

A single captive-bred, red-footed tortoise (G. carbonaria) par-

ticipated in this study. She (formerly referred to by Wilkinson

et al. 2007 as he) was approximately 4 years old and her plas-

tron (the lower part of the shell) measured 10 cm in length at

the start of the experiment. The tortoise, named Moses, had

previously been the subject of a similar spatial learning exper-

iment (Wilkinson et al. 2007). A single animal was used in

this experiment because it is hard to purchase animals older

than hatchlings in the UK. In the Wrst years of life hatchlings

of this species remain under leaf litter and move about very

little. Moses was purchased by the Wrst author, from a breeder,

when she was young. When not involved in an experimental

session, Moses was kept in a 61 cm 30 cm 30 cm glass

tank in an oYce adjacent to the experimental room. The oYce

was kept on a daily 12L:12D cycle (light on 08:0020:00).

The temperature within the tank was maintained at 29C

(4C); humidity was maintained at 50%. Moses had access

to food (fruit and vegetables) for 60 min each day; this was

given approximately 30 min after the completion of the exper-

imental session.

Apparatus

The apparatus was an eight-arm radial maze made of

opaque black Perspex. Each arm was 18.5 cm long and the

width at the arm entrance was 10 cm, and widened through-

out the arm to reach 25 cm at the end; this allowed the tor-

toise room to turn. The side walls of the maze were 10 cm

high. All arms radiated out of a central octagonal platform,

this measured 25 cm in diameter. During training and

experimental trials white plastic bottle tops (measuring

3 cm in diameter and 1.5 cm high) were placed at the end of

each arm and used as food cups. The lip of the cup pre-vented Moses from seeing whether food was available until

she had fully entered the arm.

The maze was positioned in a small experimental room

that was lit by two 60-W ceiling lights and maintained

approximately 29C by an electric heater in the corner of

the room. A circular black curtain (diameter 84 cm) sur-

rounded the maze. The length of the curtain extended

86 cm above the maze and obscured all visual cues from

the surrounding room. A large circle of card (diameter also

84 cm) was Wtted directly above the curtain; this obscured

possible ceiling cues. The circular card had a central hole

(26 cm diameter) and eight smaller holes (8 cm diameter)

placed at regular intervals around the circumference, which

provided light to the experimental set-up.

To observe Moses movements in the maze, an NDS-27

Panasonic video camera was positioned directly above the

maze (through the central circle in the card). The underside

of the camera (excluding the lens) and beams were covered

by plain brown card to prevent them being used as a cue.

The camera was connected directly to a video screen in the

experimental room; this allowed the experimenter (who

remained in the experimental room throughout the experi-

mental session) to view the tortoises movements without

interfering with the experimental set up.

Four brightly colored shapes were pinned to the curtain to

provide potential cues for navigation. They were positioned

53 cm (measured from the Xoor of the maze to the center of

the stimulus) above the maze and set at 90 intervals around

the circular curtain. Their position on the curtain was at an

equidistant point between two arms. This ensured that no

single cue could be used to detect whether a speciWc arm

had been visited, but a combination of at least two cues was

required. The cues used were a yellow triangle, blue circle,

red square (all 256 cm2), and a green cross (231 cm2).

Procedure

The experiment was run over a period of 8 weeks from July

2007 to September 2007. All trials took place in the after-

noon between 13:00 and 16:00, as this was the time that

Moses was most active. Prior to each experimental session

Moses was removed from her tank and handled for approx-

imately 5 min. This increased her activity level. She was

then placed into a wooden holding cage and transported to

the experimental room.


4/9

782 Anim Cogn (2009) 12:779787

123

Moses was given 3 days of pretraining to ensure that she

moved conWdently around and ate readily in the maze. Each

pretraining session consisted of a single trial lasting 30 min,

in which the tortoise was placed in the maze and allowed to

explore. The tortoise had access to all of the arms through-

out the experiment. On the Wrst day of pretraining, food

(dandelions and strawberries) was scattered throughout the

maze. For the following 2 days food was scattered only inthe eight arms. By the third day Moses readily entered and

ate from all eight arms within the 30 min trial period.

On each of the next 23 days of training, Moses received

four trials separated by an intertrial interval of 5 min that

was spent in the holding cage. Between each trial the maze

was rotated by either 45, 90, or 180 to discourage the

tortoise from using intramaze cues in conjunction with

extramaze cues. During each trial, the experimenter

observed and documented the tortoises behavior via the

video monitor. Entering an arm that had not been entered

previously was scored as a correct response. At the end of

each daily session the maze was wiped clean; it was not

cleaned between trials.

In the Wrst phase of training there were 15 daily sessions.

At the start of each trial Moses was placed in the central

area facing a randomly selected arm. Each food-cup was

baited with a piece of strawberry (approximately 2 mm

wide and 3 mm in length). The experimenter recorded

every arm that Moses entered. She remained in the maze

until she had entered all eight arms or 30 min had elapsed.

An arm choice was recorded when half of the tortoise had

entered the arm (although Moses rarely backed out of an

arm after entering that far). After 7 days of training, there

were no obvious signs of learning which raised the concern

that the position of the cues might be such that they were

not easily visible to the subject. At this point, therefore, the

vertical position of the cues on the curtain was changed.

They were Wxed in a position so that the bottom of each was

in line with the top of the maze wall (and 10 cm above the

level of the maze Xoor). Moses received training with this

arrangement for 8 days.

At this stage a further procedural change was made in

the hope of increasing the rapidity of learning. When train-

ing rats in a radial arm maze it is common practice to

restrict the number of choices allowed on a given trial (e.g.,

Olton and Samuelson 1976). For the next phase of training

we restricted Moses to eight choices per trial, removing her

from the maze at this point (or after 30 min).

After performing consistently above chance in the sec-ond phase of training in the controlled cue environment we

tested the extent to which Moses used the extramaze cues.

She received two test sessions. Each was made up of two

test trials intermixed with two retraining trials. The Wrst and

third trials of each session were the same as in the training

sessions, but in the second and fourth trials the cues were

removed. Moses performance was scored in the same man-

ner as for a training trial. If performance was based on

using any (or all) of the cues we would expect a total dis-

ruption of performance in the test trials.

Results and discussion

For the Wrst 15 sessions of training the subject was allowed

as many choices as were needed to complete the maze

(enter all eight arms). We scored, for each trial, the number

of correct choices in the Wrst eight. (On 14 of the 60 trials of

this phase, the subject did not complete the maze within the

30 min allowed; for these we scored the total number of

correct responses made). Interestingly, performance on the

incomplete trials was rarely less than on equivalent com-

pleted ones. Figure 1a presents the daily mean scores.

According to Olton (1978), the number of correct responses

in the Wrst eight choices to be expected on the basis of

chance, is 5.3. (Chance performance is computed assuming

that every choice is made at random, without replacement).

As the Wgure shows, mean daily scores diVered little from

the chance expectation. The mean number of correct

responses in the Wrst eight over all days of this stage was

5.68; a one-sample t test comparing this score against

chance expectation revealed that it was approaching, but

Fig. 1 a Correct choices in the

Wrst eight trials during training inthe Wrst phase of the controlled

cue environment. b Correct

choices in the Wrst eight trials

during training in the second

phase of the controlled cue

environment

0

1

2

3

4

5

6

7

8

0 2 4 6 8 10 12 14 16

Session

Correctchoicesinthefirsteig

ht

0

1

2

3

4

5

6

7

8

0 1 2 3 4 5 6 7 8

Session

Correctchoicesinthefirsteig

ht

a b


5/9

Anim Cogn (2009) 12:779787 783

123

did not reach a signiWcant diVerence, t14 = 2.05, P = 0.06.

As the Wgure shows, there was no trend toward an improve-

ment over the phase; a separate test of performance over the

last eight trials (for which the mean correct score in the Wrst

eight choices was 5.88) again revealed no diVerence from

chance, t7 = 1.44, P = 0.19.

Figure 1b shows performance over the second phase of

training, in which Moses was allowed only eight choicesper trial. She successfully completed a total of 31 out of 32

possible trials in this phase. As the Wgure shows, perfor-

mance was improved under this training regime. The mean

number of correct choices per trial over all 8 days was

6.34, a score that diVered from chance t7 = 4.72, P = 0.002.

The mean score for the last eight trials was 6.63 which

again diVered from chance expectation t7 = 5.04,

P = 0.002.

Table 1 shows the number of correct choices made on

each trial of the test phase. It is evident that removing the

cues produced no disruption of performance. Moses scored

on average 7.25 (out of 8) on the test trials and 7 on the

intermixed retraining trials. A one-sample t test revealed

that both these measures were signiWcantly above chance

(5.3), t3 = 4.07, P = 0.03 and t3 = 4.16, P = 0.03, respec-

tively. Furthermore, a paired sample test revealed that they

did not diVer signiWcantly from each other t3 = 0.52,

P = 0.64. This indicates that Moses was not using the avail-

able visual cues to navigate the maze.

Examination of the details of her performance within a

given trial revealed that Moses behavior appeared to be

determined by a simple, but highly eVective strategy, that

of turning in a given direction and entering the arm next to

the one she had just left. For each of the trials of the second

phase of training we scored the number of consecutive

choices, after the Wrst that were both in the same direction

as the previous choice and were to an arm that was adjacent

to it. Figure 2 shows these scores (mean values of the four

trials in a day). The Wgure shows that the mean length of

such a run increased dramatically over training; initially,

the tendency to turn into an adjacent arm was rather low,

but by the end of training the animal was regularly showing

runs of four or Wve such turns. Analysis of the training data

(comparing the Wrst four and last four sessions) revealed a

signiWcant increase in the choice of arms that were adjacent

to the arm that had just been left, and were in the same

direction as the previous choice t3 = 5.16, P = 0.01. Further

analyses revealed a signiWcant correlation between the

mean number of correct choices per trial and the use of the

turn-by-one-arm strategy r= 0.58, n = 32, P = 0.001. This

suggests that this behavior did lead to greater success in the

maze. Of the 32 trials in this phase of training, Moses was

as likely to turn in a clockwise direction (61 choices) as in

an anticlockwise direction (45 choices), t30 = 0.73,

P = 0.47.

This response strategy was maintained on the test ses-

sion. For the four test trials with the cues removed, the

mean run-length score was 5.5 and the mean score on the

intermixed training trials was 4.5.

In summary, Moses performed well in the radial arm

maze and scored above chance for the entire second phase

of training when the procedure limited her to eight choices

per trial. This behavior was the result of a simple response-

based strategy. This strategy was independent of the extra-

maze cues as it was maintained during the test in which

these were removed. Interestingly, she did not learn a turn-

right, or turn-left rule; although within a trial she turned

consistently in one direction, the choice of direction varied

between trials. Making a choice that is one arm away from

the previous choice is unusual in rats (Yoerg and Kamil

1982). This is likely to be due to the constraints imposed by

the tight turning circle. These constraints will no doubt

apply to the tortoise as well as the rat (perhaps more so), in

which case, the strategy adopted by Moses by the end of

training was acquired in spite of constraints that would tend

to oppose it.

Table 1 Correct choices for each trial of the cue-use test

There were two test days each containing two retraining and two test

trials. Maximum score was 8

Retraining trials Test trials

7 8

7 6

6 7

8 8

Fig. 2 The number of consecutive one arm turns moving in the same

direction in the second phase of the controlled cue environment and

the test

0

1

2

3

4

5

0 1 2 3 4 5 6 7 8 9 10

Session

Thenumberof

consecutive1arm

turnsinthe

samedirection

Training Test


6/9

784 Anim Cogn (2009) 12:779787

123

In the previous study of Moses maze behavior (Wilkin-

son et al. 2007) we found above-chance performance, but

no indication that this was based on a response strategy.

Exhaustive analysis of response sequences revealed only a

slight tendency to preferentially choose arms that were two

away from that being exited. There was no tendency to

choose adjacent arms. The diVerence between the results of

the two studies must be assumed to lie in some feature ofthe diVering procedures. The most obvious diVerence is that

Wilkinson et al. (2007) gave access to a rich visual environ-

ment (the room). In the present experiment the only extra-

maze cues available were those attached to the curtain

surrounding the maze. Although tortoises have good color

vision and are able to distinguish the colors that were used

as cues (Quaranta 1952) the positioning of the cues with

respect to the maze and to each other, or the overall sparse-

ness of the environment (not only were there were less fea-

tures available, but there was also no geometric information

provided by the room) may have prevented the tortoise

from using these cues successfully. The use of a response-

based strategy may have been dictated by the unavailability

of appropriate extramaze cues.

Experiment 2

The Wndings of Experiment 1 present a stark contrast to

those of our previous experiment (Wilkinson et al. 2007).

We have suggested that the degraded nature of the visual

cues in Experiment 1 may have prevented Moses from

using them to navigate, and that the use of the turn-by-one-

arm strategy was a result of this. If this were the case then it

is possible that, were a full range of extramaze cues made

available, Moses might stop using the response-based strat-

egy and revert to using the extramaze cues to navigate. It is

also possible, of course, that the development of the

response strategy seen in Experiment 1 was the result of the

tortoise having had extended experience in the maze.

In order to distinguish between these possibilities we

conducted a further study in which Moses was trained in a

cue-rich environment that was essentially identical to that

of the previous study in which she exhibited no turn-by-

one-arm strategy (Wilkinson et al. 2007). If the former

hypothesis is correct then we would expect the use of the

turn-by-one-arm strategy to decrease when the tortoise has

access to the full array of cues from the room. However, if

the latter is correct then we would expect no change in

behavior.

Method

This experiment began immediately after completion of the

test phase in Experiment 1. The same subject was used in

this experiment. All housing conditions remained the same.

The fabric curtain was removed from the maze so that

Moses was able to see the entire experimental room

(described below). All other apparatus remained the same.

This set up was almost identical to that used by Wilkinson

et al. (2007). External cues that were, in principle, visible

from inside the maze included shelving upon which labora-

tory equipment was stored, a large poster, the experimenter,and a black door (for exact details please see Wilkinson

et al. 2007). The experimenter could now observe the tor-

toise directly, but continued to make behavioral observa-

tions via the television screen to ensure consistency over

the two experiments. In all other ways, the procedure was

the same as described for the second phase of training for

Experiment 1. Moses received four trials per session. There

were 14 daily sessions of training and Moses completed 53

out of the 56 trials.

Results and discussion

Figure 3 shows the mean number of correct choices per

trial for each training session in the cue rich environment.

It is evident that performance remained above chance

throughout; comparison of the overall mean for each ses-

sion with chance expectation showed a signiWcant diVer-

ence, t13 = 6.64, P < 0.001. But analysis of the detailed

pattern of responding suggests that the basis of the perfor-

mance changed over the course of training. As for Experi-

ment 1 we analyzed, for each trial, the number of

consecutive selections of an adjacent arm in the same

direction of movement. Mean daily scores are shown in

Fig. 4. It is clear that use of the turn-by-one-arm strategy

declined across days; comparing the mean score for the

Wrst four and last four sessions revealed a signiWcant

Fig. 3 Correct choices in the Wrst eight trials in the cue rich training

environment

0

1

2

3

4

5

6

7

8

0 2 4 6 8 10 12 14

Session

Correctchoices

inthefirsteight


7/9

Anim Cogn (2009) 12:779787 785

123

diVerence, t3 = 7.13, P = 0.006. Further analysis revealed

that in the Wrst four sessions accuracy correlated with the

use of the turn-by-one-arm strategy r= 0.63, n = 16,

P = 0.009; however, in the last four sessions when use of

this strategy was reduced to chance levels, success no

longer signiWcantly correlated with strategy use r= 0.33,

n = 16, P = 0.22. In fact, by the Wnal session of training

the animals mean run length for turns into the adjacent

arm was just 0.75; nonetheless, overall navigational accu-

racy remained above chance.

This change in performance indicates that the response-

based strategy developed in Experiment 1 was a conse-

quence of the particular cue environment used in that study;

when the full set of extramaze cues was made available the

animal reverted to behavior like that previously seen in the

study by Wilkinson et al. (2007). It is not clear what ele-

ments of the visual environment are relevant to the change

in behavior. It is possible that the turn-by-one-arm strategy

was a response to the general poverty of the visual features

within the environment; however, the introduction of the

curtain also resulted in a change in the geometric informa-

tion available. Geometric information has been shown to be

a powerful environmental cue for animal spatial navigation

(e.g., Cheng 1986, for a recent review see Cheng 2008).

This experiment cannot clarify which environmental cues

contributed to the tortoises change in behavior; however,

what is clear is the intriguing Xexibility of behavior dis-

played by the tortoise. Navigation on the basis of the visual

environment appears to be in some sense the preferred

option; but when appropriate visual cues are not available,

the animal adopts a simple and eVective response-based

strategy. When visual cues are made available this strategy

is abandoned.

General discussion

Wilkinson et al. (2007) demonstrated eYcient radial maze

performance in a tortoise. There was no evidence of a

response-based strategy; rather the behavior appeared to be

controlled by extramaze cues, paralleling that shown by rats

in the same situation. The original aim of the present study

was to examine directly the role of extramaze cues by train-ing the tortoise in an apparatus with a restricted range of

cues that could be manipulated by the experimenter. This

aim could not be fulWlled as, when trained in an impover-

ished cue environment in Experiment 1, Moses learned to

use a simple but eYcient strategy to navigate through the

mazeto turn consistently in one direction, selecting the

arm adjacent to the one that she had just left. In Experiment

2, when the curtain was removed and the cue environment

was comparatively rich (both in terms of features and the

geometric information available), the use of the turn-by-

one-arm strategy signiWcantly decreased. Accurate perfor-

mance was nonetheless maintained. This suggests the

presence of two navigational processes in the tortoise. The

use of visual cues appears to be a default behavior which is

used if it is possible for the tortoise to do so. The second is

a response-based strategy that allowed the tortoise to navi-

gate the maze in a highly successful manner. It seems that

by surrounding the maze with a curtain and presenting only

four distal visual cues in Experiment 1, the amount of

visual information available was reduced to a level that pre-

vented the tortoise from using the extramaze cues to navi-

gate.

This pattern of behavior is quite diVerent from that

shown by rats in a similar situation. Rats in a cue-impover-

ished environment (e.g., Mazmanian and Roberts 1983) and

even blinded rats (e.g., Zoladek and Roberts 1978) do not

normally exhibit such response-based behavior. An excep-

tion is found in studies by Einon (1980) and Roberts and

Dale (1981); both experiments observed response-based

strategies in a cue-rich environment, and neither saw the

dramatic changes in the behavior that were observed in our

tortoise. However, in the latter experiment (Roberts and

Dale 1981) the eVects of proactive interference between tri-

als reduced over time, suggesting an improvement in the

use of the response-based strategy. Einon (1980) systemati-

cally examined the cause of the response-based behavior

and found that it was only present in immature rats. This

prompts the hypothesis that the strategy observed in the

maze may be related to changes in foraging behavior over

the life cycle. The nutritional requirements of immature rats

(those required for growth) are quite diVerent from what is

needed to maintain health at maturity (Clarke 1980). The

diet of wild-living rats (Rattus rattus), reXects this and

because of natural variation in the distributions of food

types, wild foraging behavior does change with age (Clarke

Fig. 4 The number of consecutive one arm turns moving in the same

direction in the cue rich environment

0

1

2

3

4

5

0 2 4 6 8 10 12 14

Session

Thenumberof

consecutive1arm

turnsinthe

samedirection


8/9

786 Anim Cogn (2009) 12:779787

123

1980). But whatever is true of rats in this sort of analysis

cannot account for our Wndings in the tortoise. Moses

search behavior changed because of cue availability (or her

ability to use the cues that were available); it was not a

result of changes in her development, suggesting that the

processes involved in her search behavior diVer from those

observed in rats (and other mammals).

This response-based navigational behavior has also beenobserved in Wsh (Roitblat et al. 1982; Hughes and Blight

1999). Like the rat studies (Einon 1980; Roberts and Dale

1981), the former study did not take place under limited cue

conditions. However, the latter study tested Wsh in both a

cue-rich and cue-impoverished environment. The aim was to

directly test the role of behavioral ecology in spatial search

behavior. The authors used two species ofWsh (Spinachia

spinachia and Crenilabrus melops) that inhabit highly struc-

tured tidal zones in which visual cues are provided by rocks

and weeds. However, rough weather frequently results in

these features being obscured. Thus the authors predicted

that the Wsh would be able to use both response-based and

spatial cue-based navigational strategies.

Their experiment goes some way to supporting our data.

They found that in a cue-rich environment the Wsh were

able to use cues to navigate; however, in the impoverished

cue environment the Wsh used a response-based strategy.

Interestingly, both species were signiWcantly less good in

the non-cue than in the cue condition; this is quite diVerent

from Moses behavior. Furthermore, their experiment was

run between subjects, so it is not possible to know whether

the Wsh would change from a successful response-based

strategy to a cue-based strategy with a change in the cue

environment. The authors interpreted their results as being

due to the behavioral ecology of the two species that were

tested (both lived in tidal zones). However, to our knowl-

edge these speciWc tests have not been given to non-tidal

species ofWsh, thus, it is possible that this behavior is com-

mon to all Wsh (and maybe reptiles).

The small amount of literature available on spatial

behavior in reptiles does not provide any previous observa-

tions of the animals using a response-based strategy to nav-

igate. Such evidence as there is for the use of multiple

memory strategies in chelonian comes from the study of

Lpez et al. (2001) who found that terrapins were able to

use both a salient (beacon) cue and a map-like representa-

tion to navigate to a single goal. The animals in this set of

experiments, however, were tested in a between-subjects

design, which did not allow examination of the possibility

of changes between these strategies in a single animal. In

contrast, in our Experiment 2 we found that Moses showed

a change of navigational strategy according to environmen-

tal circumstances. It is interesting that she did not need to

change her behavior to successfully navigate the maze in

the cue rich environment; continuing with the turn-by-one-

arm strategy would have maintained a high level of perfor-

mance. The only comparable example of such Xexibility

can be seen in the lesioned terrapins of Lpez et al. (2003)

that had to re-learn to navigate to a goal post-operatively.

The lesions appeared to prevent them from using the map-

like representation that was observed in the training phase

(and in the sham-operated animals), and thus they were

forced to learn a beacon-type strategy to reach the goal. It isunclear, however, why the changes in Moses navigational

behavior occurred. Our results suggest that the use of visual

cues is a default strategy that will be overruled when eVec-

tive visual cues are not available. It is not clear whether the

same is true of the Wsh tested by Hughes and Blight (1999);

however, there is a clear similarity in behavior between

their Wsh and our tortoise.

In sum, these Wndings suggest the presence of two pro-

cesses that can control navigation in our tortoise. One

appears to be based on visual cues. It is apparently similar

to the cognitive-map mechanism employed by rats; but our

attempt to investigate this directly in Experiment 1 was

thwarted by the animals use of a diVerent mechanism. This

second mechanism involves a simple response-based strat-

egy of a sort not usually observed in mammals, but appears

to be present in at least some Wsh. This pattern ofWndings

suggest that when tortoises (or at least this tortoise) navi-

gate in a situation with poor environmental cues they use a

simple, but eYcient response-based strategy; but when

more cues are available they switch away from this, and

apparently navigate using the surrounding visual cues. This

behavioral Xexibility may reXect the behavioral ecology of

the speciWc species tested, or it may be true of chelonia and

possibly reptiles and Wsh in general.

Acknowledgment The participation of Sacha Coward was made

possible by a bursary from the UK Experimental Psychology Society.

The authors would like to thank Paul McLaughlin and Simon Fletcher

for making the maze. We are also indebted to the Behavioural Neuro-

science Research Group at the University of York and the Biology of

Cognition Research Group at the University of Vienna for their helpful

comments.

References

Balda R, Kamil A (2006) The ecology and life history of seed cachingcorvids. In: Brown MF, Cook RG (eds) Animal spatial cognition:

comparative, neural, and computational approaches. Available

on-line at: www.pigeon.psy.tufts.edu/asc/balda/

Broglio C, Gomez A, Duran E, Ocana F, Jimenez-Moya Rodriguez F,

Salas C (2005) Hallmarks of a common forebrain vertebrate plan:

specialized pallial areas for spatial, temporal and emotional mem-

ory in actinopterygian Wsh. Brain Res Bull 66:277281

Burghardt GM (1977) Learning processes in reptiles. In: Gans C,

Tinkle DW (eds) Biology of the Reptilia. Academic, London, pp

555681

Cheng K (1986) A purely geometric module in the rats spatial repre-

sentation. Cognition 23:149178
http://www.pigeon.psy.tufts.edu/asc/balda/http://www.pigeon.psy.tufts.edu/asc/balda/


9/9

Anim Cogn (2009) 12:779787 787

123

Cheng K (2008) Whither geometry? Troubles of the geometric mod-

ule. Trends Cogn Sci 12:355361

Clarke DA (1980) Age- and sex-dependent foraging strategies of a

small mammalian omnivore. J Anim Ecol 49:549563

Day LB, Crews D, Wilczynski W (1999) Spatial and reversal learning

in congeneric lizards with diVerent foraging strategies. Anim

Behav 57:393407

Day LB, Crews D, Wilczynski W (2001) EVects of medial and dorsal

cortex lessons on spatial memory in lizards. Behav Brain Res

118:2742

Einon D (1980) Spatial memory and response strategies in rats: age,

sex and rearing diVerences in performance. Q J Exp Psychol

32:473489

Hart DR (1983) Dietary and habitat shift with size of red-eared turtle

(Pseudemys scripta). Herpetologica 39:285290

Holtzman DA, Harris TW, Aranguren G, Bostocks E (1999) Spatial

learning of an escape task by young corn snakes (Elaphe guttata

guttata). Anim Behav 57:5160

Hughes RN, Blight CM (1999) Algorithmic behaviour and spatial

memory are used by intertidal Wsh species to solve the radial

maze. Anim Behav 58:601613

Kamil AC, Balda RP, Olson DJ (1994) Performance of four seed cach-

ing corvid species in the radial-arm maze analog. J Comp Psychol

108:385393Lpez JC, Rodrguez F, Gmez Y, Vargas JP, Broglio C, Salas C

(2000) Place and cue learning in turtles. Anim Learn Behav

28:360372

Lpez JC, Gmez Y, Rodrguez F, Broglio C, Vargas JP, Salas C

(2001) Spatial learning in turtles. Anim Cogn 4:4959

Lpez JC, Vargas JP, Gmez Y, Salas C (2003) Spatial and non-spatial

learning in turtles: the role of the medial cortex. Behav Brain Res

143:109120

MacPhail EM (1982) Brain and intelligence in vertebrates. Clarendon

Press, Oxford

Mazmanian DS, Roberts WA (1983) Spatial memory in rats under

restricted viewing conditions. Learn Motiv 14:123139

OKeefe J, Nadel L (1978) The hippocampus as a cognitive map.

Clarendon Press, Oxford

Olton DS (1978) Characteristics of spatial memory. In: Hulse SH,

Fowler H, Honig WK (eds) Cognitive processes in animal behav-

ior. Erlbaum, Hillsdale, pp 341373

Olton DS, Samuelson RJ (1976) Remembrance of places passed: spa-

tial memory in rats. J Exp Psychol Anim Behav Process 2:97116

Quaranta JV (1952) An experimental study of color vision of the giant

tortoise. Zoologica 37:295314

Roberts WA, Dale RH (1981) Remembrance of places lasts. Proactive

inhibition and patterns of choice in rat spatial memory. Learn

Motiv 12:261281

Roitblat HL, Tham W, Golub L (1982) Performance ofBetta splendens

in a radial arm maze. Anim Learn Behav 10:108114

Strong JN (2005) Seed dispersal and the ecological implications of

hunting Geochelone carbonaria and G. denticulate in Northwest-

ern Brazil. Masters thesis, College of Environmental Science and

Forestry, State University of New York, New York

Strong JN, Fragoso JMV (2006) Seed dispersal by Geochelone

carbonaria and Geochelone denticulate in Northwestern Brazil.

Biotropica 38:683686

Tinklepaugh OL (1932) Maze learning of a turtle. J Comp Psychol13:201206

Wilkinson A, Chan H, Hall G (2007) Spatial learning and memory in

the tortoise (Geochelone carbonaria). J Comp Psychol 121:412

418

Yoerg SI, Kamil AC (1982) Response strategies in the radial arm

maze: running around in circles. Anim Learn Behav 10:530534

Zoladek L, Roberts WA (1978) The sensory basis of spatial memory in

the rat. Anim Learn Behav 6:7781

Documents

Navegacion en Tortugas